Genetic analysis and multimodal imaging confirm m.12148 T > C mitochondrial variant pathogenicity leading to multisystem dysfunction

1. Introduction

Mitochondrial transfer RNA (mt-tRNA) variants significantly cause mitochondrial DNA (mtDNA) disorders [1,2]. Variations of mitochondrial transfer RNAs notably occur more frequently than in proteincoding segments of the mitochondrial genome [3]. The higher frequency of mt-tRNA variants is believed to result from the stem-loop structure formed during mitochondrial DNA replication, which can expose single-stranded portions of mt-tRNA genes to higher oxidative damage rates than that of other mtDNA regions. [4,5] However, despite the high frequency of pathogenic variants in mt-tRNA genes, most mt-tRNA variants are benign or likely benign [3]. A single abnormal mt-tRNA has the potential to alter the translation of multiple mtDNA-encoded proteins, leading to an increased likelihood of metabolic disruptions by influencing events across the entire electron transport chain (ETC).

Determining whether an mtDNA variant is disease-causing is a challenging clinical task, partly due to the variability in heteroplasmy between cells and/or tissues [4]. This also means that mtDNA variants in a heteroplasmic state might be missed during the clinical diagnostic process, as tissues carrying mtDNA variants and experiencing symptoms may not be readily accessible for biopsy, such as muscular, cardiac, nervous, and renal systems. Clinical assessments of heteroplasmy are often performed using blood and buccal swabs, which do not accurately reflect variant heteroplasmy in affected organs [4,5]. Furthermore, the most affected tissues in these diseases, neural and cardiac tissues, are typically unavailable for in vitro functional studies. The emergence of induced pluripotent stem cells (iPSCs) as a tool to model tissues provides a method to address some of these challenges, thereby facilitating the study of the functional impact of particular mtDNA variants on tissues that are otherwise difficult to access.

Herein, we describe a mother and daughter harboring the m.12148 T > C variant,first described in Wong et al 2020, in the mt-t RNA for histidine (mt-tRNA^His) gene (Eurasian population frequency 2.4 × 10⁻³; total population frequency³ 1.6 × 10⁻³) [3] with mitochondrial syndrome affecting multipe organ systems. Although this cohort’s data has been previously entered into the MITOMAP database, this variant is currently classified as “likely pathogenic.” Through a comprehensive characterization of this specific mt-tRNA syndrome using fibroblasts and iPSC-differentiated cardiac and neuronal cells, we demonstrated the effects of this mt-tRNA variant on mitochondrial morphology and cell function. Therefore, we provide a definitive characterization of the disease affecting this family and verify that the m.12148 T > C variant is the source of mitochondrial dysfunction. Our study supports the classification of this variant as pathogenic and provides a precedent for monitoring patients for potential mitochondrial diseases when the m.12148 T > C variant is detected.

2. Materials and methods

This study was approved by the Institutional Review Board of Stanford University. All participants provided written informed consent before undergoing evaluation, testing, and sequencing.

2.1. Cell line establishment

To establish fibroblast lines, fresh 3 mm skin punch biopsies were dipped in 70 % ethanol and then cut into smaller pieces prior to seeding onto gelatin-coated 6-well tissue culture plates with a cover slip placed on top to aid in attachment. Skin punches were then cultured in DMEM supplemented with 10 % FBS and antibiotic-antimycotics (Gibco, 15,240,096) for the first five days, then without antibiotics. Fibroblast growth was observed at the edges of the punch biopsies on day 14, at which point the cells were trypsinized and seeded into new gelatin-coated 6 well plates for subsequent experimental expansion, sequencing, and/or cryopreservation in Bambanker (Lymphotec Inc.). All fibroblasts were maintained for six passages. For heteroplasmy measurements in other cell types, specifically blood and urine-collected cells, the cells were clinically captured and used for genetic testing only.

2.2. Single cell sort for outgrowth and heteroplasmy determination

To initiate clonal expansion of our lines and determine single-cell heteroplasmy, cells were sorted using a BD Biosciences FACS ARIA III based on scatter and then seeded at one cell per well into a 96-well PCR or cell culture plate. To determine heteroplasmy at the single cell level, cells were sorted into PCR plates and heated to 100 °C, mitochondrial genomes were amplified using REPLI-g Mitochondrial DNA Kit (Qiagen, 151,023), and the resulting material was tested for heteroplasmy using qPCR. Cells plated into 96 well culture plates were expanded from the single-cell level until confluency and then passaged into 12- or 24-well plates for further expansion, experimental use, or for heteroplasmy qPCR assessment. Lines that consistently showed zero heteroplasmy were designated as wild type (WT), and lines that consistently showed >60 % heteroplasmy were designated as heteroplasmic. Image and biochemical analyses, as well as iPSC reprogramming of these isogenic fibroblast lines, which differed only in their mitochondrial heteroplasmy status, were used for subsequent experiments.

2.3. qPCR determination of heteroplasmy

Amplification-refractory-mutation-system (ARMS) qPCR primers were designed to selectively amplify the variant mtDNA to determine heteroplasmy. Two pairs of sequence-specific primers were designed with penultimate mismatches and base changes to decrease hairpin formation and primer dimerization [8]. Multiple primer sets were assayed with the following being selected for use: COMMON FWD 5’-AACACAATGGGGCTCACT-3,’ WT REV 5’-GACTCACGATCTGATGTTTTGGTTATA-3,’ and m12148C REV5’-GACTCACGATCTGATGTT TTGGTTATG-3.’ Primers were validated using WT DNA, plasmids carrying WT and m.12148 T > C DNA, and whole genomic DNA from control and patient samples. The primer products were visualized on agarose gels to confirm the specific amplification. Heteroplasmic quantification was confirmed by Sanger sequencing. The samples to be assayed were prepared in replicate pairs, and each qPCR assay was performed in triplicate with the required controls using BIORAD iTaq Universal SYBR Green Supermix (Bio-Rad, 1,725,120). Cycle threshold (Ct) values were then used to calculate the percentage of heteroplasmy using a previously established method: 1/ [1 + (1/2) ^ (CtWT-CtMut)] × 100 [9].

2.4. Analysis of mitochondrial morphology with artificial intelligence (AI) machine learning

Images of wild-type fibroblasts and heteroplasmic mtDNA 12,148 T > C fibroblasts (> 60 % heteroplasmy) were captured by confocal microscopy and analyzed. The high-resolution images were 1900 × 1900 pixels and showed single cells that were fixed in 4 % PFA in PBS for 15 min, permeabilized, and stained with DAPI to reveal the nuclei. To reveal the mitochondrial networks and cell membranes, cells were exposed to baculoviruses expressing fluorescent proteins (Cell Light C10600 and C10608, Thermo Fisher Scientific), which labeled the mitochondria and cell membranes, respectively, prior to fixation and permeabilization. We followed an image-based profiling methodology to analyze the sets of images [6,7]. We then quantified the mitochondrial structures of all single cells by cropping overlapping regions of 300 × 300 pixels in the areas of the images where the mitochondria had the highest activation. This resulted in approximately 32 overlapping regions from which the mitochondrial structures were sampled for quantification. In total, 418 mitochondrial regions were extracted from the images. The morphological features of each region were computed using a pre-trained convolutional neural network [6,7], resulting in a 1536-dimensional vector representing the structural content of the image. Before feature computation, the mitochondrial channel was normalized using contrast-limited adaptive histogram equalization (CLAHE). DeepProfiler was used to compute deep learning features. The feature extraction procedure resulted in a matrix of 418 regions, with 1536 features, each representing the regions extracted from the images.

2.5. Biochemical analysis of mitochondrial enzymatic activity

Complex I and IV activities were measured using the Complex I Enzyme Activity Microplate Assay Kit (ab10972, Abcam) and the Complex IV Human Specific Activity Microplate Assay Kit (ab109910, Abcam). For complex I measurements, fibroblasts were homogenized, and proteins were detergent-extracted and quantified using standard biochemical assay methods (23,225, Pierce BCA Protein Assay Kit). Extracts of lines and appropriate controls were incubated for 3 h in the appropriate buffers. The assay solution was added, and the OD at 450 nm was measured every 30 s for 30 min. The rate and slope were calculated for all samples using the following equation: Rate mOD/min = Abs 1 – Abs 2 (mOD at @ 450 nm) Time 1 – Time 2 (min).

For complex IV measurements, fibroblasts were homogenized, and proteins were detergent-extracted and quantified using standard BCA methods. Samples and appropriate controls were loaded onto plates and incubated for 3 h. Assay solutions were then added, and the OD at 550 nm was measured at 1-min intervals for 2 h to determine enzyme activity: Rate mOD/min = Abs 1 – Abs 2 (mOD @ 550 nm) Time 1 – Time 2 (min). To determine the total amount of bound complex protein in the plate, antibodies against complex IV were added to the wells used in the enzyme activity assay. The bound protein was reacted according to kit parameters and detected via OD measurements at 405 nm every minute for 30 min. The bound complex concentration relative to the total loaded protein was calculated within the linear portion of the curve and used to normalize complex IV enzyme activity.

2.6. Analysis of mitochondrial respiration

Mitochondrial respiration was assayed using a Seahorse XFe96 Extracellular Flux Analyzer (Agilent Technologies). Neural-derived lineages were measured on day 45 post-differentiation to coincide with synapsin-positive cells within the culture and cardiac lineages were measured on days 55–60 post-differentiation based on culture health, density, and beating. The derived cell lines were plated on XF96 cell culture microplates and allowed to grow for 48 h. Prior to the assay, the cells were equilibrated in XF DMEM supplemented with 25 mM glucose, 2 mM Glutamax, and 1 mM pyruvate (103680–100, Agilent Technologies). Respiration parameters were then assayed using the Mito Stress Test Kit (103708–100, Agilent Technologies) and by applying oligomycin (1.5 μM), carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP, 2 μM) and rotenone/antimycin A (0.5 μM). Basal respiration rates, maximal respiration (respiration in response to FCCP) and reserve capacity (maximal respiration – basal respiration) were then established.

2.7. Establishment of isogenic induced pluripotent stem cell (iPSC) lines

iPSC lines derived from isogenic fibroblast cells were created as per standard protocol using CytoTune^™-iPS 2.0 Sendai Reprogramming Kit (A1657). Cells were seeded, transduced with the SENDAI virus, and grown with mTESR e8. On day 14 post-transduction, single colonies were selected and grown. After 5–7 days, cells were reseeded on feeder-free substrates and propagated to passage 20 for all experiments. WT-iPSCs maintained low heteroplasmy of m.12148 T > C and m.12148 T > C maintained approximately 60 % heteroplasmy until passage 20 and beyond.

2.8. Differentiation of dopaminergic neurons

Dopaminergic neuronal cultures were initiated by the formation of neural aggregates from isogenic iPSC lines. Dissociated single cells were loaded into AggreWell^™800 to form neural aggregates. At this stage, the cells were treated with the neural induction medium. Dual-SMAD inhibition was initiated with Noggin (Peprotech 120-10C)/SB431542, and purmorphamine (Peprotech 4,831,086) was added to specify the midbrain lineage. On day 5, the aggregates were plated onto laminin substrates and expanded with bFGF (Peprotech 100-18B), forksolin, and SHH/FGF8 (Peprotech 100-25A) to further specify the midbrain lineage. Three weeks after the start of this protocol, cultures were re-plated onto PDL/LAMININ in a neurobasal N2/B27 media with FGF8, SHH, TGFβ-3 (Peprotech 100-36E), BDNF (Peprotech 450–02), GDNF (Peprotech 450–10), forksolin, and ascorbic acid. FGF8 and bFGF were removed from the culture on day 27, and SHH levels were reduced. Cells were fed half media changes every other day and cultured for extended periods by replating onto fresh PDL/laminin.

2.9. Differentiation of cardiomyocytes

Cardiomyocytes were derived using the metabolic maturation protocol as previously described [8]. Confluent lines were cultured in RPMI (Gibco 11,875,101) supplemented with B27 (without insulin) plus 6 μM CHIR. CHIR concentrations were reduced by the addition of media over the following 2 days. On day 3, the media was changed and supplemented with 2 uM Wnt–C59. On day 7, the medium was changed to RPMI supplemented with B27 and insulin was added. Subsequently, the medium was changed every other day until day 11. Cells were then placed in a starvation state (RPMI without B27 or glucose) for 3 days. On day 15, cells were replated with RPMI + B27 + insulin. On day 17, the final starvation state was induced for 3 days. On day 20, cells were cultured in RPMI+B27 + insulin and then cultured in maturation media from that point on (as described in Feyen et al.) [8].

2.10. Cardiomyocyte contractility measurements

HiPSC-CMs were plated on Geltrex at 40,000 cells per well of a 384-well plate (Greiner BioOne) in 75 μL of RPMI/B27 + insulin containing 10 % KOSR and ROCK inhibitor Y-27632 (Tocris, 1253). The next day, the medium was changed to RPMI/B27 + insulin final volume 90 μL and cells were maintained for a minimum of 7 days prior to imaging. For the analysis, 45 μL of the medium was removed, and hiPSC-CMs were loaded with 45 μL 2× TMRM (T668, ThermoFisher Scientific) with Hoechst 33258 (ThermoFisher Scientific). After incubation for 10 min in a 37 °C 5 % CO₂ incubator, cells were washed by removing 90 μL of the medium and replacing it with 90 μL RPMI/B27 + insulin. After a 2-h recovery period, timed series images were acquired automatically using the IC200 KIC instrument (Vala Sciences, California, USA) at an acquisition frequency of 50 Hz for a duration of 10 s. Contractility image analysis using TMRM signal-derived vectorized movement points through a time series and physiological parameter calculations were conducted as previously described [16]. The D-peak represents the peak divergence or the average peak of contraction across a 10-s recording, graphically representing contractile changes relative to controls [9].

3. Results

3.1. Patient details

The proband underwent a genetic evaluation at 12 years of age for bilateral progressive sensorineural hearing loss and early cortical cataracts. Over the years, she developed retinopathy, ataxia, tremors, declining cognitive function, absence of seizures, frequent headaches, fatigue, mild proximal muscle weakness, autonomic dysfunction, and renal disease. Muscle biopsy revealed cytochrome oxidase (COX)-deficient fibers (Figs. 1A–B).

Fig. 1.

Multiple tissues demonstrating functional evidence of mitochondrial disease. Histopathological and immunohistochemical studies in the patient’s muscle. (A) Cytochrome c oxidase (COX) histochemistry (i), succinate dehydrogenase (SDH) histochemistry (ii) and sequential COX-SDH histochemistry (iii) demonstrate a mosaic pattern of COX deficiency in the patient’s muscle, with several fibers showing intermediate COX activity. (B) Quadruple immunofluorescence analysis of NDUFB8 (complex I) and COXI (complex IV) protein expression in the patient’s muscle. Each dot represents the measurement of an individual muscle fiber, colour coordinated according to its mitochondrial mass (low = blue, normal = beige, high = orange, very high = red). Gray dashed lines represent SD limits for the classification of the fibers. Lines next to the x- and y- axis represent the levels of NDUFB8 and COXI: beige = normal (> −3), light beige = intermediate positive (−3 to −4.5), light purple = intermediate negative (−4.5 to −6), purple = deficient (< −6). Bold dashed lines represent the mean expression level of normal fibers. These data confirm the loss of COX1 expression in several fibers, consistent with the molecular genetic defect identified and the COX-SDH histochemistry. (C) Head CT scan at the age of 12 years revealing significant basal ganglia and cerebellar calcifications. (D) Muscle biopsy specimens from proband acquired at the age of 12 years showing haematoxylin and eosin (H&E) staining of muscle, including increased nicotinamide adenine dinucleotide dehydrogenase (NADH) staining (red arrows), increased trichome staining (blue arrows), and increased succinate dehydrogenase (SDH) staining, all indicative of mitochondrial disease. (E) Samples acquired the same year were examined by transmission electron microscopy (TEM), which showed abnormal mitochondrial morphology (red arrows) in muscle with two representative sections (zoomed out-top and zoomed in-bottom panel). (F) At the age of 21 years, samples from a renal biopsy were acquired. H&E staining presented collapsed glomeruli, with antibody staining showing increased complement 3 (C3) deposits and Immunoglobulin M (IgM) staining revealing inflammation. (G) TEM imaging of renal tissue, revealing dysmorphic features and abnormal mitochondria morphology. (H) Multimodal retinal imaging in both eyes reveals symmetrical waxy disc pallor (top panel), significant loss of retinal pigment epithelial (RPE) cells on fundus autofluorescence (FAF) (middle panel), and outer retinal atrophy with loss of the ellipsoid zone (EZ) and interdigitation zone (IZ) on optical coherence tomography (OCT) (bottom panel). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Computed tomography (CT) scanning of the brain revealed intracranial calcifications (Fig. 1C). This was further evaluated using magnetic resonance imaging (MRI), which revealed abnormal hyperintense T1 signals correlating with areas of abnormal calcification noted on the CT scan, including the globus pallidi, putamina, dorsal thalami, caudate, and dentate nuclei. On T2 weighted images, the caudate and putamen demonstrate an abnormally increased signal. Subsequent MRI examinations were not possible owing to the presence of cochlear implants. Further investigation revealed morphologically abnormal mitochondria (Fig. 1D), which were visualized by electron microscopy (Fig. 1E). Anatomical pathological evaluation included trichrome staining, which revealed some fibers with subsarcolemmal granular staining but no ragged red fibers. Electron microscopy also revealed an increased number of mitochondria, some of which were abnormally shaped, but no characteristic intramitochondrial inclusions were observed.

Sequencing of mtDNA performed at the Baylor College of Medicine on the initial muscle biopsy from the age of 12 years revealed a heter-oplasmic (% unspecified) variant of uncertain significance: m.12148 T > C in the mt-tRNA for histidine. In her third decade, she developed renal disease in the form of focal segmental glomerular sclerosis (FSGS) associated with mitochondrial abnormalities (Figs. 1F–G). Her retinopathy also progressed (Fig. 1H), with the loss of retinal pigment epithelial cells (RPE) and retinal atrophy. The aforementioned renal disease and progressive renal failure were contributing factors to her death, along with a stroke-like episode a week before her death at the age of 30. Additional details about her clinical course and family history are available in the Supplementary Section.

The proband’s mother had later onset and less severe bilateral hearing loss, and cataracts developed during her 5th decade of life. Comprehensive mtDNA sequencing was performed at the Baylor College of Medicine, revealing that the m.12148 T > C variant was present at 10 % heteroplasmy in the urine of the proband’s mother. Subsequent mtDNA sequencing at Stanford University detected the m.12148 T > C variant in the mother, with 1 % heteroplasmy in the fibroblasts and 5 % heteroplasmy in the blood. The unaffected younger sister, now of childbearing age, was evaluated by PCR-based mtDNA sequencing for the m.12148 T > C variant, which was undetectable in a urine sample as detection was limited to >1 % heteroplasmy. Similarly, a buccal swab of the unaffected maternal grandmother tested negative for this variant. Whole Exome Sequencing (WES) performed for the proband, mother, or sister identified no confounding nuclear variants. (Supplementary Table 2).

3.2. Bulk tissue and single cell analysis of mitochondrial DNA heteroplasmy

Considering the presence of the likely pathogenic m.12148 T > C variant, we evaluated heteroplasmy in tissues from the proband, including blood, dermal fibroblasts, and urine sedimentary cells (obtained five years premortem). Heteroplasmy was determined using an allele refractory mutation system (ARMS)-based quantitative PCR and was reflected as the percentage of amplified m.12148 T > C copies relative to wild-type mitochondrial DNA. This analysis revealed that all samples harbored the m.12148 T > C variant with varying levels of heteroplasmy: fibroblasts, approximately 56 %; blood, 40 %; and renal epithelial cells, 68 % (Table 1). These values were confirmed by nextgeneration sequencing of mtDNA.

Table 1.

Heteroplasmy Across Patient Tissues.

To understand the distribution of heteroplasmy in the sampled tissues, we performed single-cell heteroplasmy analysis. This was accomplished by disassociating cells carrying the m.12148 T > C variant, sorting one cell per well in a 96-well plate, and assaying the m.12148 T > C mtDNA load via ARMS-based quantitative PCR in single cells. This approach revealed that fibroblasts have a greater range of heteroplasmy than urine and blood cells (Table 1). The broad range of heteroplasmy observed in the fibroblast lines made it a useful tool for subsequent investigations, in which we clonally expanded the isogenic wild-type (WT) and heteroplasmic lines (HP) of fibroblasts for functional testing.

3.3. Image analysis of mitochondrial morphology with varying levels of heteroplasmy using artificial intelligence (AI) machine learning

The relationship between heteroplasmy for m.12148 T > C and mitochondrial morphology was investigated in fibroblast lines that had the greatest variance in heteroplasmy, comparing highly heteroplasmic (HP) lines (>60 %) to those with no detectable variant or “wild-type” (WT) lines. Observation of fluorescently labeled mitochondria in the HP and WT lines highlighted differences in mitochondrial networks (Fig. 2A). Cell images were partitioned into mitochondrial regions (red boxes) to increase the experimental n from 13 to 418, thereby increasing statistical significance (Fig. 2B). Subsequently, machine learning was applied to identify differences between the two cell lines. A convolutional neural network (CNN) was trained to measure the visual differences in natural images to characterize mitochondrial morphology. Rather than targeting specific structures or phenotypes, this method can objectively and quantitatively detect the emergent properties of mitochondria. Principal component analysis (PCA) was used to reduce the dimensionality of the features from 1536 to 100 dimensions (Fig. 2C). In this reduced feature space, we performed multiple-hypothesis corrected Student’s t-tests to identify which of these dimensions was relevant for distinguishing between HP and WT fibroblasts. When plotting significantly different dimensions, we observed a consistent separation of mitochondrial patterns between groups (Fig. 2C), suggesting that mitochondrial morphology was affected by the presence of the m.12148 T > C variant. Finally, we tested whether the distribution of correlations between mitochondrial regions could be organized into three different groups: self-correlation of the wild type, self-correlation of heteroplasmic regions, and cross-correlation between the wild type and heteroplasmic regions. We anticipated that the self-correlations would have high values and that the cross-correlations would have low values, meaning that there were consistent intra-group similarities and significant inter-group differences. To test this hypothesis, we built a distribution of correlations between the three groups and used a Kruskal–Wallis nonparametric test, assuming the null hypothesis that no difference exists between the distributions. The results are shown in Fig. 2D, indicating that the differences observed between the WT and HP mitochondrial regions were statistically significant (p < 0.0001).

Fig. 2.

Mitochondrial Imaging. (A) High-resolution confocal images of fibroblasts included in this analysis, showing normal fibroblasts and fibroblasts with >60 % heteroplasmy (high heteroplasmic-HP) fibroblasts. (B) Regions of the mitochondria channel extracted and analyzed in this study. The original resolution of confocal images is 1900 × 1900 pixels (left), and an overlapping region of 300 × 300 pixels was automatically identified in the mitochondrial channel to compute morphological features. The image is first split in a regular grid, and regions with mito activity higher than a threshold are kept for quantification. (C) Principal component analysis of the morphological features of mitochondrial regions in the 13 images of cells. The morphological features are a set of 1536 latent variables obtained with a pre-trained convolutional neural network (CNN) to assess visual variations in images. Each point represents a 300 × 300-pixel region in the mito channel. Points in blue are mito regions from wild-type fibroblasts, while points in red are mito regions from heteroplasmic fibroblasts. The principal components 4, 5, 6 and 9 were significantly different between the two populations of cells and are shown in this fig. (D) Distribution of correlations between regions and single cells in three groups: wild-type self-correlation (WT-WT), heteroplasmic self-correlation (HP-HP), and wild-type heteroplasmic cross-correlation (WT-HP). The distribution of similarities at the single-cell level was evaluated. The blue (WT-WT) and green (HP-HP) distributions have tails to the right that indicate differences with respect to the orange (WT-HP) distribution. These differences were non-significant. However, the region-level correlations had statistically significant differences. The green distribution (HP-HP) presents shifts in both tails with respect to the orange distribution (WT-HP), indicating a difference between heteroplasmic mito and wild-type regions. The blue distribution (WT-WT) peaks significantly at zero and the right tail, highlighting the differences regarding cross-correlations (WT-HP). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

By contrast, the intercellular results were not statistically significant at the single-cell level. This lack of statistical power at the single-cell level is likely due to the stochastic nature of heteroplasmy, which makes it difficult to parse morphological differences. However, intracellular mitochondrial region analysis potentially enhances the statistical power by highlighting subcellular morphological differences in the mitochondria when individual cells have both wild-type and variant mtDNA.

3.4. Biochemical analysis of mitochondrial enzymatic activity and mitochondrial respiration

To determine whether the differences in mitochondrial morphology in heteroplasmic m.12148 T > C-carrying cells were also correlated with mitochondrial function, we analyzed mitochondrial respiration and enzyme activity. The activities of complex I and IV were assayed because they contained a very high concentration of histidine, which could reflect altered mt-tRNA^His function in m.12148 T > C carriers. To assess the activities of complex I and IV, four lines were derived directly from proband fibroblasts, two with zero or near zero variant burden and two lines with a high heteroplasmic load of m.12148 T > C (60–80 %). These lines were clonally expanded from single cells. Complexes were isolated from whole-cell extracts and normalized to total protein, and enzymatic activity was assessed by ELISA. Heteroplasmic lines carrying m.12148 T > C showed significantly less complex I activity than the isogenic control lines (p = 0.0030; ** one-way ANOVA) (Fig. 3A). This significant reduction in enzyme activity was also observed when assaying complex IV (P < 0.0001 ****, one-way ANOVA). (Fig. 3B). Lines with high variant loads had significantly lower complex IV activity than the isogenic control lines. The isogenic design of this experiment, requisite protein normalization, and the above-mentioned results confirmed that m.12148 T > C tRNA^His had a detrimental effect on ETC kinetics in isolated experiments.

Fig. 3.

Biochemical and mitochondrial analysis of WT and m.12148 T > C cell lines highlights deficits in the activities of complex I and IV and cell respiration. (A) Enzymatic activity of complex I in clonally derived patient fibroblast lines with determined heteroplasmy (Homo (−) 0 %, and Hetero (+/−) +60 %) homoplasmic lines had higher activity than heteroplasmic lines, (p = 0.0030 ** one-way ANOVA). (B) Complex IV activity in clonally derived patient fibroblast lines with determined heteroplasmy Control (−) 0 % and Hetero (+/−) +60 %) were expanded and harvested to determine complex IV activity. Homoplasmic lines had higher activity than heteroplasmic lines (P value <0.0001 P value summary **** one-way ANOVA). (C) Clonally derived fibroblast lines seeded for the Seahorse mitochondrial stress test, heteroplasmic lines showed lower respiration in all measured categories compared to the control unpaired t-test p-value <0.0001 p-value summary ****. (D) Single experiment representative of the mitochondrial stress test.

To investigate the metabolic implications of the m.12148 T > C tRNA^His variant, mitochondrial stress tests were performed on clonally expanded fibroblast lines using the Agilent Seahorse platform. Assaying oxygen consumption rates (OCR) yielded changes in respiratory capacities, with heteroplasmic lines having the lowest basal respiratory capacity, 57 % of the wild-type line. The measured OCR from ATP production in heteroplasmic lines ranged from 43 % to 66 % relative to isogenic WT cells. Additionally, the maximum respiratory capacity in these heteroplasmic (HP) lines was 50–60 % that of the WT (Fig. 3C–D). These findings suggest that the m.12148 T > C variant impacts mitochondrial respiration and provides further evidence for its potential pathogenic impact on carriers.

3.5. Patient-derived inducible pluripotent stem cell (iPSC) lines recapitulate phenotype in affected and unaffected cell types

In addition to studying fibroblasts, we also created IPCSs from the probands to differentiate cells towards neuronal and cardiac lineages and determined the effects of the m.12148 T > C heteroplasmy on disease-relevant cell function (Fig. 4A, B). Differentiation towards these lineages reduced variant heteroplasmy to<50 % (Fig. 4C). However, we assayed cells for oxygen consumption rates to determine if these levels have a functional impact on neuronal- and cardiac-derived cells, particularly when compared to isogenic control-derived cells. This showed that neuron-derived lines differentiated by 45 days, with heteroplasmy at approximately 42.3 % (Fig. 4B), presented with a generalized reduction in oxygen consumption rates in mitochondrial stress test assays (Fig. 4D–E). In contrast, in the 50–55-day cardiomyocyte differentiations, with heteroplasmy at approximately 46.1 %, only minor basal respiration decreases were evident in the mitochondrial stress test assays (Fig. 4F–G), and no effects were detected in contractility functional assays (Fig. 4H). However, whether greater heteroplasmic levels of m.12148 T > C would eventually introduce larger phenotypic defects in cardiomyocytes is unclear. These data suggest that m.12148 T > C sensitizes neuronal phenotypic responses to cardiomyocyte responses at comparable heteroplasmic levels, which could indicate the events found in the proband with evident neurological deficits present in the absence of cardiac pathology.

Fig. 4.

IPSC-derived lines decrease in heteroplasmy after passaging and differentiation. (A) IPSC lines derived from clonal fibroblast lines yielded pluripotent cells with varying heteroplasmy. (B) Neuronal and cardiac lineages. (C) Heteroplasmic IPSC lines decrease below the disease threshold upon long-term differentiation. Blue arrows represent the timepoint for Seahorse assays: 45 days for neural lines and 55–60 days for caridiac lines. (D) Cellular respiration in dopaminergic neuronal (DAns) lineages shows lower respiration in heteroplasmic lines (approximately 42.3 % HETEROPLASMY) in all categories measured: Basal Respiration ATP production, Maximal Respiration and Spare Capacity. Unpaired t-test P value 0.0011 P value summary **Significantly different. (E) Single experiment representative of the mitochondrial stress test in DAns. (F) Cellular respiration in patient-derived cardiomyocytes (CMs) shows respiration is lower in heteroplasmic lines (approximately 46.1 % HETEROPLASMY) only in Basal Respiration Unpaired t-test P value 0.0307 P value summary *Significantly different (P < 0.05). (G) Single experimental representative graph of mitochondrial stress. (H) Contractility of isogenic lines shows no significant differences in contractility (Student t-test) p > 0.05. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4. Discussion

The m.12148 T > C variant has only been reported from this single family, initially as a variant of uncertain significance (VUS) and later as likely pathogenic when the variant was identified in the proband’s mother [3]. The current study adds to previous studies of this variant in this isolated family, providing additional clinical and genetic analysis of the proband and mother of the proband, including functional evidence and morphological analysis on the pathogenic nature of the m.12148 T > C. Ultimately, our data support the reclassification of this variant from likely pathogenic to pathogenic. To support our conclusion, we focused on the m.12148 T > C variant, which is located in the D arm stem loop of mt-tRNA for histidine. This single-nucleotide modification of the reference mitochondrial DNA genome is presumed to disrupt base pairing of the stem loop structure and potentially alter aminoacyl-tRNA synthetase activity [10,11]. However, this has not yet been proven to disrupt mitochondrial function. We provided expanded data on this variant and presented a mother and daughter carrying the m.12148 T > C variant with optic, aural, renal, and neurologic symptoms, along with evidence of its functional impact on relevant patient-specific cell types and biochemical assays.

This study demonstrated mitochondrial dysfunction in tissue biopsies and quantified heteroplasmy for m.12148 T > C in samples from different tissues, including blood, skin, and urine of the proband and her mother. As expected, heteroplasmy was confirmed to be distinctly different across blood, skin, and urine samples from the daughter (proband), who presented with the most severe disease state, as measured by clinical assessment of brain, eye, muscle, and kidney biopsies and/or imaging studies. The mother, who presented with the disease after the proband’s diagnosis, had previously tested negative in blood and positive in urine for the m.12148 T > C mutation and was positive for heteroplasmic levels of m.12148 T > C in blood, urine, and skin with hearing loss and cognitive decline. We further found that the heteroplasmy of m.12148 T > C correlated with distinct differences in mitochondrial morphology and functional deficits in ETC enzyme activity and respiration in proband-captured fibroblasts, as well as in diseaserelevant IPSC-derived neurons.

Furthermore, by assessing single-cell heteroplasmy, we observed that different cells within tissues had different levels of m.12148 T > C heteroplasmy, a finding previously masked by bulk-cell approaches. Specifically, peripheral blood monocytes and bladder epithelial cells in the urine had a tight distribution of heteroplasmy, whereas fibroblast cells showed a large variation in heteroplasmy, ranging from a small number of cells with no detectable variant to a population that was largely homoplasmic for the m.12148 T > C variant. This wide variation in fibroblast heteroplasmy is potentially reflective of both the degree of severity and the diversity of neurological symptoms found in our proband, considering that fibroblasts and neurons are both embryologically derived from the ectoderm [11–13]. The diversity of the heteroplasmic load in fibroblasts is also likely reflective of their mitotic activity, as previous studies have identified an asymmetric distribution of mitochondrial content during cell division [14]. However, this heteroplasmic diversity in fibroblasts allowed us to clonally expand the cells towards creating lines stably expressing m.12148 T > C at 60 % and undetectable levels (isogenic control fibroblasts).

As a metric to confirm clinical mitochondrial morphology alterations, we used these clonally expanded fibroblast lines to provide strong evidence for morphological alterations by m.12148 T > C heteroplasmy, with the added benefit of measuring changes against control fibroblasts. Notably, although historical measures of mitochondrial morphology alteration have aimed to assess the size, number, distribution, and cristae organization of mitochondria, the present study aimed to provide a novel approach for automated unbiased measures of mitochondrial change. The approach was informed by 1536 variables and was based on pre-trained variations from mitochondrial-marked images in healthy and disease-carrying cells. Ultimately, this provides an image-based indicator for mitochondrial dysfunction in m.12148 T > C-carrying cells and is reflective of prior investigations in which known mtDNA variants alter mitochondrial morphology [15].

Following image-based alterations in the mitochondria, we aimed to determine whether these events coalesced with functional alterations in m.12148 T > C-carrying cells. When comparing mitochondrial respiration and complex I and IV enzyme functions between isogenic control and variant-carrying fibroblast lines, m.12148 T > C-carrying fibroblasts exhibited significantly compromised respiratory and ETC enzyme activity. Thus suggesting the pathogenic nature of the m.12148 T > C variant and presenting the first evidence for the functional impact of the m.12148 T > C variant. Furthermore, to study the effects of the m.12148 T > C variant on patient-relevant tissues, particularly those often affected by mtDNA variants, we generated IPSC lines carrying clonally expanded fibroblasts with and without detectable m.12148 T > C heteroplasmy. We consider this method as an alternative to the traditional use of cybrid lines, which can be cumbersome and has certain limitations, such as the possible presence of confounding nuclear gene variants [16,17,29]. Most importantly, using iPSCs allowed us to create cell types that represent tissues with similar metabolic demands as those expressing disease symptoms within our proband (neural), as well as tissues commonly associated with mitochondrial disease (cardiac).

Specifically, we aimed to identify neurons relevant to the proband who presented with loss of motor function and calcification in the basil ganglia, suggesting the involvement of dopaminergic neurons. To further highlight the importance of these neurons in the functional assessment of the m.12148 T > C variant, other mtDNA variants of known pathogenicity have been historically linked and tested for their contribution to the substantia nigra and dopaminergic neuron degeneration and dysfunction, such as in MELAS-associated parkinsonism’s [18–20]. Similarly, cardiac cells have been studied because they are often involved in mitochondrial diseases, such as MELAS and MERFFS [21,22]. Ultimately, our testing in iPSC-derived dopaminergic neurons carrying the approximately 40 m.12148 T > C variant yielded evident decreases in respiratory capacity compared with ionogenic control neurons. No detectable changes in cardiac cell respiration or contractility were noted at similar heteroplasmy levels. These observations could be attributed to the nature of this variant and the cardiac tissues, as our proband did not present with any cardiac-related symptoms. However, the level of heteroplasmy in the cardiac tissue was possibly below the necessary threshold for pathogenicity in our proband, and this effect was also below the threshold in our cardiac-differentiated iPSC lines. Other studies have suggested that higher heteroplasmy levels are required to observe cardiac phenotypes, as in MELAS studies. Furthermore, post-mitotic tissues such as cardiac cells and neurons can increase heteroplasmy with aging [23]; thus, cardiac involvement might show as the proband’s mother ages (the proband is deceased). In differentiating iPSCs into neurons and cardiomyocytes, heteroplasmy levels decreased relative to the initial differentiation days, possibly because of mitotic selection during the cell progenitor stages against m.12148 T > C carrying mitochondria or even from unknown selective pressure as cells entered post-mitotic cell fates as cardiomyocytes and neurons. Future studies should focus on understanding the selection criteria for mtDNA variants in neurons and cardiomyocytes to understand the selection for and against specific variants.

Overall, we found that in all cases, compared to isogenic control cells, whether fibroblasts or iPSC-derived neurons, m.12148 T > C cells carrying the variant experienced diminished ETC enzyme activity and respiration capacities. Our findings, combined with clinical workups, multi-tissue biochemical analysis, genetic analysis, and imaging data, indicate that m.12148 T > C has the potential to alter normal mitochondrial function. Furthermore, layering these findings onto the detection of this variant only in the subjects experiencing clinically impactful and deleterious symptoms within our cohort supports the reclassification of m.12148 T > C from a “likely pathogenic” mtDNA variant to a “confirmed pathogenic” mtDNA variant. However, this does not necessarily mean that m.12148 T > C is definitively causative of a particular disease state when detected in a patient; rather, it cannot be excluded as a source of a potential disease when detected.

Notably, the immediately adjacent m.12147 base pair, which also codes for tRNA^His, has exhibited a pathogenic variant m.12147 A > G (G12147A) present in a previously healthy young man who had an abrupt onset of symptoms manifesting as a severe stroke-like episode at the age of 22 years [24]. This contrasts with the gradual development of symptoms observed in our proband. The proximity of these two variants, both of which code for tRNA^His, and each exhibiting pathological variants causing severe yet different symptoms, supports the significance and vulnerability of this mtDNA region.

Determining whether an mtDNA variant causes disease is complex and largely depends on the heteroplasmic state of the variant. The reason heteroplasmy disseminates differently across tissues and within individual cell types is not entirely understood. Multiple factors may be at play, including different developmental selection events [25–28]. These events may involve different metabolic and growth pressures driving alternate levels of variants within tissues, which are further modified by cellular heterogeneity within tissue types. For example, dermal fibroblast-derived cells contain a mixture of heterogeneous cell types that move between states of rapid migration/growth [11–13], lending a wide distribution of heteroplasmic states, as observed in our results. Exposure to environmental factors in utero or post-natally may increase or decrease heteroplasmy in an individual, explaining notable differences in heteroplasmy and disease manifestation across and within generations [2]. Although the reasons for heteroplasmic differences across cell lineages are not fully understood, these phenomena must be considered in the clinical assessment of mitochondrial disease. A high variant load in a small but relevant cell population may be masked in bulk sequencing runs of non-relevant cell populations, leading to falsenegative results when mitochondrial disease is suspected.

This study demonstrated the comprehensive characterization of a specific mitochondrial transfer RNA variant, underscoring the importance of multiple tissue site sampling and sequencing with single-cell analysis. Additionally, we demonstrated a method for quantifying heteroplasmy at the single-cell level and single-cell clonal expansion to generate isogenic iPSCs that reflect in vivo disease characteristics. The combination of the data from molecular and cellular investigations, including the clinical correlation with familial studies, provides evidence for reclassification of the m.12148 T > C mt-tRNA^His variant from “likely pathogenic” to “pathogenic,” thus adding to the increasing list of confirmed pathogenic mitochondrial transfer RNA variants in the public database, which may prove useful in genetic counseling for future patients. Additionally, our techniques havepotential applications in future mitochondria-targeted therapeutics and/or clinical trials.

Data availability

Data will be made available on request.